PerceptionHardware: Max~3 min training per scene

Edge NeRF capture

Neural Radiance Fields for environment capture, optimized for Jetson.

About this demo

Capture a real environment by walking the robot around it; the brain trains a NeRF on-device that becomes the basis for a real-to-sim scene. Inspired by Johnny Núñez Cano's Edge NeRF & nerfstudio talk.

Highlights

→On-device training
→nerfstudio integration
→Exports to Real-to-Sim pipeline
→Mesh + radiance field outputs

Supported robots

Mobile capture rig

Related demos

View all →

Perception

Real-time object detection (YOLOv11)

TensorRT INT8 detection on dual RGBD streams with 3D position.

Open demo →Teleop

CockPit

Predefined teleop dashboard with dual cameras, 3D model, map, and controls.

Open demo →Dashboard

My UI

Drag-and-drop dashboard — every CockPit widget, your layout.

Open demo →